Vector quantization (VQ) is a simple and useful data compression algorithm which has been widely applied in many fields such as image processing and pattern recognition. Because each data block is encoded by only one approximate vector in the codebook, the accuracy of the reconstructed blocks is usually poor in VQ. In this paper, the bilinear vector quantization (BVQ) algorithm is proposed with simple expressions. Experiment results show that the quality of the reconstructed data blocks in BVQ is much better than that in VQ when their encoding time is kept the same. Hence, BVQ is more efficient and more effective than VQ.